Rule-based transport mode classification in a smart travel and mobility survey

Smartphones offer an innovative way to collect survey data on travel behavior through smart travel surveys. Smart travel apps potentially reduce the response burden, aiming to become fully automated and requiring no respondent interaction. To explore this, Statistics Netherlands developed a smart-travel app and collected data from a sample of the Dutch population. This paper focuses on classifying the mode of transport, a central smart functionality of the app, by developing a deterministic rule-based algorithm to classify seven transport modes: bike, bus, car, metro, train, tram, and walking. Three manual rule-based algorithms were developed that differed in the feature sets used. The final algorithm achieved an overall classification accuracy of 85% and a balanced accuracy of 70%. Public transport modes are the most challenging to classify. The results show that combining GPS and OSM data is important for achieving transport mode classifications of acceptable quality. The developed algorithm shows that a manual rule-based approach can be highly effective in classifying transport modes. Manual rule-based approaches thus offer several advantages over machine-learning approaches and should not be overlooked.

Fourie, J., J. Klingwort, Y. Gootzen (2025). Rule‐based transport mode classification in a smart travel and mobility survey. Discussion paper, Statistics Netherlands, The Hague/Heerlen.